Progress 09/01/13 to 04/30/15
Outputs
Target Audience: Target Audience - Our audience was largely produce packaing companies, known as "packing sheds" wherin produce affected by PTI compliance is packed. These are lagely in California, as that is where 75% of the US's fresh produce comes from. Within that market segment we were looking at sheds that produce a variety of produce and would need active identification and coding of produce cases with PTI compliant lot, date and shed information. Efforts - Our efforts involved partnering with an existing equipment supplier to build out and test a "proof of concept" prototype using machine vision and ink jet priting to evaluate the feasibility a "whole product" solution wherin the system can identify a produce case by visual parent and child markings and then apply PTI compliant GS-128 barcodes using an inkjet printer in a simulated production envoronment. Changes/Problems: Because of the intricasy of the CODE 128 bar code - there up to 32 characters being represented in a bar code - material handling plays a very important challenge in producing quality, repeatable bar PTI compliance markings. Ink jet printing is by far the best choice for the producer because of the lower purchase and operating costs, however, it has to be coupled with a material handling solution. In phase I we proved that we could identify cases and contents and then print CODE 128 bar codes. The remainder of the work has to do with making those markings 100% grade A readable bar codes. What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest? Our results have been disseminated to our partners and shared within potential target customer groups. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
Impact - With PTI compliance, produce is safer. It is safer because trace-backs, a method of determining where a retail level produce originated from can be done in less than a day where prior to the initiative it took several weeks sorting through a paper trail at each step between producer and retailer. This allows for faster recalls, which translates into fewer deaths and pinpoint disposal of affected goods instead of wholesale category destruction (less loss to supplier and retailer). TECHNICAL OBJECTIVES Our Phase I grant was awarded in support of Executive Order 13329 “Encouraging Innovation in Manufacturing”. Our award was to support the coding and marking of electronic traceback information on produce packaging lines. Pursuant to this objective, we broke our Phase I project into three distinct technical objectives. Objective 1 – Identify GTIN products based on outer case markings using machine vision, Objective 2 – Imprint electronic traceability information (trace back) directly on the case using inkjet technology, and Objective 3 - determine the optimal solution. Specifically, from our proposal we asked the following questions regarding our objectives. Objective 1 – Machine Vision · Can we differentiate between produce sku’s based on outer case markings and/or dynamics of the case? · How fast can we discriminate between sku’s? · How difficult is it to train the vision system? Objective 2 – Printing Traceability Information · Are we able to stabilize the box in order to print realizable code 128 barcodes? · Can we adapt to the physics of produce cases in real time during a production run? · What are the best ink sets for this application? Objective 3 – Optimal Solution · Determine the optimal solution based on system and operational costs, reliability and accuracy with respect to inkjet, labels and laser coding. TESTING OBJECTIVE I We found a solution that was capable of identifying different cases and contents via case markings that was fully repeatable and able to work up to 180 feet per minute, the maximum observed conveyor line speed. We researched solutions for machine vision on the market and partnered with Product Jet a company in Fresno, California that manufacturers equipment for the produce industry that we felt could be incorporated into a system to meet our end objective of a whole product solution for identifying and coding produce cases. IDENTIFYING DIFFERENT SKU'S The machine vision system works by comparing images captured on an assembly line to images stored in a library. It analyzes both invariant and variant patterns - invariant patterns being fixed, pre-printed markings and the variant patterns being hand stamped or written markings (invariant and variant markings have a parent-child relationship). Invariant patterns determine the case/GTIN, and then variant patterns are used to determine case contents. For example, an invariant marking determines that a case is a box of oranges, then a child search (under the parent) returns the specific size of oranges. Per the figure below, the vision system can be trained to recogonize parent and child (invariant and variant) features and markings on a produce case. This allowed us to identify the case and contents, the first step in identifying and then printing PTI compliant information. Figure - Image Software Trained to Identify a Produce Case TESTING OBJECTIVE II We were able to stabilize the box and print CODE 128 barcodes on our system test bed, which lays the foundation for Phase II work which will focus on material handling solutions and surface preparation. We built three test beds. A laboratory test bed, that is has precise material handling indicative of a laboratory environment, and then two additional test beds intended to test material handling concepts for precise material handling in production environments. On our system test bed, Figure 5, we were able to repeatedly produce Grade A CODE 128 barcodes. However, our production test beds, Figure 10, produced barcodes did not adequately support material for repeatable Grade A CODE 128 bar codes. Our production test beds were designed to tightly control the product as it passed through the print area by forcing the product up against the print head from the back, and top, and to precisely control the horizontal movement by providing horizontal travel via drive motors and belts – our solution was intended to be placed in series with an existing production line, and not rely on travel from the line itself, as many production line belts provide non-linear movement. This approach was designed to take out as many variables as possible. We believed we could restrict and control product movement and allow for less movement between the case and the print head, while testing revealed it had the opposite effect. In order to get the cases close enough to the print head surface, which was protected by a large vertical skid plate, a back product guide (see red arrow, Test Bed 3) introduced enough friction that it would momentarily stop the product (for a fraction of a second – was not visible to the naked eye, but was clearly visible in output barcodes). This meant the horizontal movement was non-linear. In worst case, with cases with little weight, this effect was magnified. When additional weight was added, the effect was minimized, but could be detected using a rotary encoder or by the scan-ability and quality of the barcodes printed. Figure - Material Handling Production Test Beds 2 & 3 LESSONS LEARNED IN MATERIAL HANDLING Controlling the movement of product during the printing process presents a host of issues due largely to static friction, and its impact on linear movement. For phase II, we intend to modify the production test beds to align the product against the print surface PRIOR TO PRINTING. Note: The figure below is a very simplistic rendering used to illustrate the basic concept that product positioning is done ahead of printing. The production version will be much more complex and will require a fair degree of both mechanical and electrical engineering. Figure - Proposed Phase II Material Handling Concept TESTING OBJECTIVE III – DEFINING AN OPTIMAL SOLUTION What is the optimal solution based on system and capital and operational costs, reliability and accuracy with respect to inkjet, labels and laser coding? The ideal printing solution is to use thermal inkjet printers in conjunction with a material handling solution to solution that supports repeatable Grade A CODE 128 barcodes and a machine vision system to identify different products and contents. This is due to the significantly lower capital and operating costs compared to the only other present alternative – thermal transfer labels. Assumptions: Cases coded per year per shed – 1,250,000 Working days per season – 150 days Cases per day – 8333 Average lines per shed – 4 Table 2– Thermal Transfer vs. Inkjet Direct Printing Thermal Transfer Labels Inkjet Printing Pros Proven technology Lower operating and equipment costs. Direct print is a permanent mark. Cons Expensive to purchase and operate. Slow cycle time results in extra machine purchases Labels can become detached from cases. Grade A barcode (in production) repeatability TBD Barcode Scanability >99% TBD Reliability High High Cycle Time 1 second 0.2 second Cap X $15,800 $7,500 (with material handler) Op X $6.71/M $0.67/M Capital and Operating Coding Expense per Shed Cap X per shed $63,200 $30,000 Op X per shed $8,387 $837
Publications
|
Progress 09/01/13 to 04/30/14
Outputs
Target Audience:
Nothing Reported
Changes/Problems: PROBLEMS ENCOUNTERED The material handling stations we built, Figure 1, operated under the assumption that we would control the motion of the cases throughout the printing process. In testing with GS-1 barcodes (which use only 14 characters), we are not achieving consistent high-grade barcodes. The idea of actively controlling products and thereby product motion [during printing], in theory is sound, but in reality it seems to magnify the problems. Largely this is due to friction in the vertical skid plates (red arrow right of Figure 1). The material handling stations, in order to push the product close enough to the print head, end up causing the product to “seize up”. This is also compounded by using boxes that are not assembled properly such that they are square. At this point we will focus on printing repeatable CODE-128 barcodes on our laboratory test bed and use Phase II to focus on material handling solutions given what we have learned with our two material handling test beds. An example of the CODE-128 barcode is shown in figure 2 What opportunities for training and professional development has the project provided?
Nothing Reported
How have the results been disseminated to communities of interest?
Nothing Reported
What do you plan to do during the next reporting period to accomplish the goals? PROGRESS TOWARD ORIGIONAL WORK PLAN At this point in the project we have met our first objective in testing and qualifying a machine vision system that is capable of identifying cases and contents of produce boxes. We have set up three test fixtures to evaluate material handling for the printing of PTI compliant information which we anticipate to be challenging due to the number of characters (24-34) represented in the CODE-128 barcode portion of the PTI compliant marking. PROBLEMS ENCOUNTERED The material handling stations we built, Figure 1, operated under the assumption that we would control the motion of the cases throughout the printing process. In testing with GS-1 barcodes (which use only 14 characters), we are not achieving consistent high-grade barcodes. The idea of actively controlling products and thereby product motion [during printing], in theory is sound, but in reality it seems to magnify the problems. Largely this is due to friction in the vertical skid plates (red arrow right of Figure 1). The material handling stations, in order to push the product close enough to the print head, end up causing the product to “seize up”. This is also compounded by using boxes that are not assembled properly such that they are square. At this point we will focus on printing repeatable CODE-128 barcodes on our laboratory test bed and use Phase II to focus on material handling solutions given what we have learned with our two material handling test beds. An example of the CODE-128 barcode is shown in figure 2 SUCCESSES TO DATE We were able to achieve the first half of our goal and identify GTIN products based on outer case markings. We visited several packaging sheds in Fresno, California and made notes on issues and conditions experienced in real world production environments. We then qualified the system in-house by setting up models for several different GTIN products and evaluating the accuracy of the vision system at speeds up to the maximum observed line speed. We found that the system performed flawlessly up to the maximum speeds. Our next task is to qualify our printer under the same speed conditions and then mate the two systems.
Impacts
What was accomplished under these goals?
This was initially emailed over a few months ago. I am pasting it in here now that we know this is where we are to submit final reporting (instead of emailing components). -Matthew Brown EXECUTIVE SUMMARY This is a late delivery of our mid-term results. Our projective was to develop a machine vision system capable of identifying produce packages (GTIN items) and contents and then to mark the cases with the corresponding PTI compliant information per the PTI initiative described in our Phase I proposal. At the midpoint of the project, we succeeded in both finding and qualifying a machine vision system that could be programmed to correctly identify produce boxes from models, and their contents based on secondary box markings, i.e., a human mark in a fill in box indicating large, medium, or small oranges. We qualified the machine vision system on a conveyor system built to act as a laboratory (not production) test environment. At this point in the project, we had not qualified or tested imprinting PTI compliant information, but were ½ way to our objective.
Publications
|